TikTok and disinformation, a perfect match

To study how the electoral campaign was evolving on TikTok, we analyzed videos posted during the month of January that used the surnames of the leading presidential candidates as hashtags (Figure 1). In this snapshot, we found numerous disinformation videos among the most-viewed content.

Figure 1

For example, two videos from @actu.nofake using the hashtag #macron and posted on January 24 got 532,200 views (Figure 2) and 395,800 views (Figure 3) respectively. These videos suggested that schools were going to close starting January 31.

Figure 2

Figure 3

We now know for certain that this was not the case, but it's worth adding that this information was already in complete contradiction with the strategy of the Minister of National Education at the time it was published. Indeed, since the start of the pandemic, he had faced criticism for keeping schools open no matter what, sometimes against scientific recommendations. Through his imprecise statements about infection rates, the spread of contamination in schools was downplayed even as epidemiologists identified them as major clusters of contamination. Mr. Blanquer had also congratulated himself that France was among the OECD countries that had kept its schools closed for the shortest amount of time.

So the information spread by @actu.nofake on TikTok is absolutely false, and it runs counter to the entire strategy of the French government regarding schools. We find this same disinformation effort on Twitter, where @actu.nofake, renamed @NOFAKEactu, simultaneously spreads the same false information as on TikTok (Figure 4). However, it doesn't seem to find the same popularity there: it has only 25 followers on Twitter versus 1,489 on TikTok, and on Twitter its posts get at most 1 retweet compared to 1,467 shares on TikTok.

Figure 4

TikTok is today the most-visited website in the world and ranks fiftieth in France, so it is essential to keep an eye on this fake news and its impact during the election period. All the more so since 62% of TikTok's French users are over 17 years old (Figure 5). These potential voters risk having their political opinions skewed by false information that seems authentic simply because it has been viewed so many times.

Figure 5

Content moderation could offer a solution to this problem, but the resources social networks currently deploy are far from sufficient to tackle fake news. In 2017, TikTok had promised to reach 10,000 moderators employed, yet they have to moderate videos posted at a rate of one piece of content every 25 seconds. Within TikTok's French branch, moderation is handled mainly by an algorithm that blocks violent content and insults. It is also supplemented by human moderators who understand and speak French. One might still question the working conditions of these moderators given Candice Frazier's testimony; she worked for Telus International, a subcontractor responsible for moderating TikTok's American branch. Frazier filed a complaint over the working conditions imposed by ByteDance, TikTok's parent company. She describes 12-hour days spent watching up to 10 videos simultaneously containing acts of extreme violence: shootings, child pornography, cannibalism, animal mutilation…

So it seems unlikely that the 10,000 moderators TikTok claims to employ worldwide could verify the accuracy of the information published at such a rate, regardless of their working conditions. Unfortunately, it therefore falls to users to exercise caution and critical thinking toward the information they receive. Everyone must verify important information for themselves.